1---2name: ultrawork3description: Parallel execution engine for high-throughput task completion4---56<Purpose>7Ultrawork is a parallel execution engine and execution protocol for independent work. It emphasizes intent grounding, parallel context gathering, dependency-aware task graphs for non-trivial work, and concise evidence-backed execution summaries. It is a component, not a standalone persistence mode -- it provides parallelism and routing guidance, but not persistence, verification loops, or long-lived state management.8</Purpose>910<Use_When>11- Multiple independent tasks can run simultaneously12- User says "ulw", "ultrawork", or wants parallel execution13- You need to delegate work to multiple agents at once14- Task benefits from concurrent execution but the user will manage completion themselves15</Use_When>1617<Do_Not_Use_When>18- Task requires guaranteed completion with verification -- use `ralph` instead (ralph includes ultrawork)19- Task requires a full autonomous pipeline -- use `autopilot` instead (autopilot includes ralph which includes ultrawork)20- There is only one sequential task with no parallelism opportunity -- delegate directly to an executor agent21- User needs session persistence for resume -- use `ralph` which adds persistence on top of ultrawork22</Do_Not_Use_When>2324<Why_This_Exists>25Sequential task execution wastes time when tasks are independent. Ultrawork enables firing multiple agents simultaneously and routing each to the right model tier, reducing total execution time while controlling token costs. It is designed as a composable component that ralph and autopilot layer on top of.26</Why_This_Exists>2728<Execution_Policy>29- Fire all independent agent calls simultaneously -- never serialize independent work30- Always pass the `model` parameter explicitly when delegating31- Read `docs/shared/agent-tiers.md` before first delegation for agent selection guidance32- Use `run_in_background: true` for operations over ~30 seconds (installs, builds, tests)33- Run quick commands (git status, file reads, simple checks) in the foreground34- Resolve intent and uncertainty before implementation; explore first, ask only when still blocked35- For non-trivial tasks, produce a dependency-aware plan with parallel waves before execution36- Keep delegated-task reports concise: short summary, files touched, verification status, blockers37- Manual QA is required for implemented behavior, not just diagnostics38</Execution_Policy>3940<Steps>411. **Read agent reference**: Load `docs/shared/agent-tiers.md` for tier selection422. **Ground intent first**: Confirm whether the request is implementation, investigation, evaluation, or research; do not code before that is clear433. **Gather context in parallel**:44 - direct tools for quick reads/searches45 - exploration/docs agents for broad context464. **Classify tasks by independence**: Identify which tasks can run in parallel vs which have dependencies475. **Create a task graph for non-trivial work**:48 - Parallel Execution Waves49 - Dependency Matrix50 - acceptance criteria and verification steps per task516. **Route to correct tiers**:52 - Simple lookups/definitions: LOW tier (Haiku)53 - Standard implementation: MEDIUM tier (Sonnet)54 - Complex analysis/refactoring: HIGH tier (Opus)557. **Fire independent tasks simultaneously**: Launch all parallel-safe tasks at once568. **Run dependent tasks sequentially**: Wait for prerequisites before launching dependent work579. **Background long operations**: Builds, installs, and test suites use `run_in_background: true`5810. **Verify when all tasks complete** (lightweight):59 - Build/typecheck passes60 - Affected tests pass61 - Manual QA completed for implemented behavior62 - No new errors introduced63</Steps>6465<Tool_Usage>66- Use `Task(subagent_type="oh-my-claudecode:executor", model="haiku", ...)` for simple changes67- Use `Task(subagent_type="oh-my-claudecode:executor", model="sonnet", ...)` for standard work68- Use `Task(subagent_type="oh-my-claudecode:executor", model="opus", ...)` for complex work69- Use `run_in_background: true` for package installs, builds, and test suites70- Use foreground execution for quick status checks and file operations71</Tool_Usage>7273<Examples>74<Good>75Three independent tasks fired simultaneously:76```77Task(subagent_type="oh-my-claudecode:executor", model="haiku", prompt="Add missing type export for Config interface")78Task(subagent_type="oh-my-claudecode:executor", model="sonnet", prompt="Implement the /api/users endpoint with validation")79Task(subagent_type="oh-my-claudecode:executor", model="sonnet", prompt="Add integration tests for the auth middleware")80```81Why good: Independent tasks at appropriate tiers, all fired at once.82</Good>8384<Good>85Correct use of background execution:86```87Task(subagent_type="oh-my-claudecode:executor", model="sonnet", prompt="npm install && npm run build", run_in_background=true)88Task(subagent_type="oh-my-claudecode:executor", model="haiku", prompt="Update the README with new API endpoints")89```90Why good: Long build runs in background while short task runs in foreground.91</Good>9293<Bad>94Sequential execution of independent work:95```96result1 = Task(executor, "Add type export") # wait...97result2 = Task(executor, "Implement endpoint") # wait...98result3 = Task(executor, "Add tests") # wait...99```100Why bad: These tasks are independent. Running them sequentially wastes time.101</Bad>102103<Bad>104Wrong tier selection:105```106Task(subagent_type="oh-my-claudecode:executor", model="opus", prompt="Add a missing semicolon")107```108Why bad: Opus is expensive overkill for a trivial fix. Use executor with Haiku instead.109</Bad>110</Examples>111112<Escalation_And_Stop_Conditions>113- When ultrawork is invoked directly (not via ralph), apply lightweight verification only -- build passes, tests pass, no new errors114- For full persistence and comprehensive architect verification, recommend switching to `ralph` mode115- If a task fails repeatedly across retries, report the issue rather than retrying indefinitely116- Escalate to the user when tasks have unclear dependencies or conflicting requirements117</Escalation_And_Stop_Conditions>118119<Final_Checklist>120- [ ] All parallel tasks completed121- [ ] Build/typecheck passes122- [ ] Affected tests pass123- [ ] No new errors introduced124</Final_Checklist>125126## Parallel session caveats127128- **Multi-repo workspace anchor:** drop a `.omc-workspace` marker at the parent directory so multiple sessions across sub-repos share one `.omc/`. Resolution order: `OMC_STATE_DIR > .omc-workspace > git > cwd`. See `docs/REFERENCE.md`.129- **Session id source:** OMC_SESSION_ID env var wins in CLI contexts; hook payload data.session_id wins in hook contexts.130- **Plan id (when applicable):** Ultrawork has no persistent state; two concurrent runs are independent by design. No plan-id needed.131- **Parallel verdict:** supported (stateless component)132133<Advanced>134## Relationship to Other Modes135136```137ralph (persistence wrapper)138 \-- includes: ultrawork (this skill)139 \-- provides: parallel execution only140141autopilot (autonomous execution)142 \-- includes: ralph143 \-- includes: ultrawork (this skill)144```145146Ultrawork is the parallelism layer. Ralph adds persistence and verification. Autopilot adds the full lifecycle pipeline.147</Advanced>
salacoste